Distributions
Overview
Distributions in DESMO-J provide you with a powerful possibility to supply your simulations with a nearly endless stream of random numbers. DESMO-J uses Java's built-in random number generator (java.util.Random) which uses a linear congruential method with a cycle of 245 random numbers. The produced samples of one period can be considered statistically independent. So although computer generated random numbers are of course always deterministic (or pseudo-random), this ensures enough "randomness" to model stochastic processes.
Based on the data type of the produced samples, DESMO-J differentiates between numerical distributions, entity distributions and boolean distributions (base classes NumericalDist, EntityDist and BoolDist).. At the moment, the following distributions are ready to use:
| Class | Return type | Distribution | ||
|---|---|---|---|---|
| BoolDistBernoulli | boolean | Bernoulli distribution | ||
| BoolDistConstant | boolean | Constant boolean distribution | ||
| ContDistEmpirical | Double | Continuous empirical distribution | ||
| ContDistErlang | Double | Erlang distribution | ||
| ContDistExponential | Double | Exponential distribution | ||
| ContDistGamma | Double | Gamma distribution | ||
| ContDistNormal | Double | Normal distribution | ||
| ContDistTriangular | Double | Triangular distribution | ||
| ContDistUniform | Double | Continuous uniform distribution | ||
| DiscreteDistBinomial | Long | Binomial distribution | ||
| DiscreteDistConstant | Flexible, e.g. Integer | Constant numerical distribution | ||
| DiscreteDistEmpirical | Flexible, e.g. Integer | Discrete empirical distribution | ||
| DiscreteDistPoisson | Long | Discrete empirical distribution | ||
| DiscreteDistUniform | Long | Discrete (integer) uniform distribution | ||
| EntityDistEmpirical | Entity | Distribution of Entities (probabilities assigned per entity) | ||
| EntityDistUniform | Entity | Distribution of Entities (probabilities equal for all entities) |
The constant distributions are provided for testing purposes. Instances of these two classes always return the same value.
In addition, ContDistAggregate allows to aggregate (e.g. add or multiply) two distributions, while ContDistCustom is parameterised using an arbitrary cumulative distribution function implementing the interface desmoj.core.dist.Function.